Bug 1198005 - engine keeps vlanid as integer. Exception when using fcoe
Summary: engine keeps vlanid as integer. Exception when using fcoe
Keywords:
Status: CLOSED DUPLICATE of bug 999975
Alias: None
Product: Red Hat Enterprise Virtualization Manager
Classification: Red Hat
Component: ovirt-engine
Version: 3.4.4
Hardware: Unspecified
OS: Unspecified
high
high
Target Milestone: ---
: 3.6.0
Assignee: Lior Vernia
QA Contact:
URL:
Whiteboard: network
Depends On:
Blocks:
TreeView+ depends on / blocked
 
Reported: 2015-03-03 08:11 UTC by Pavel Zhukov
Modified: 2019-08-15 04:21 UTC (History)
14 users (show)

Fixed In Version:
Doc Type: Bug Fix
Doc Text:
Clone Of:
Environment:
Last Closed: 2015-03-05 08:26:55 UTC
oVirt Team: Network
Target Upstream Version:
Embargoed:


Attachments (Terms of Use)

Description Pavel Zhukov 2015-03-03 08:11:09 UTC
Description of problem:
If there are FCoE interfaces in the system engine throws an exception NumberFormatException in addHostVlanDevices. It tries to cast string to int but string has format "DDD-fcoe" so it fails  Auto_VLAN=no is not supported by bnx2fc so either vdsm or engine code must filter non-numeric vlans or handle it properly

Version-Release number of selected component (if applicable):


How reproducible:
100%

Steps to Reproduce:
1. Use https://access.redhat.com/solutions/1268183 to configure FCoe with bnx2fc

Actual results:
NumberFormatException

Comment 1 Yaniv Lavi 2015-03-04 12:52:04 UTC
The action item here is to update the kbase to not get this problem, can you review it and see if the steps are correct?

Comment 2 Pavel Zhukov 2015-03-04 12:57:32 UTC
(In reply to Yaniv Dary from comment #1)
> The action item here is to update the kbase to not get this problem, can you
> review it and see if the steps are correct?

We don't have workaround so far. The general workaround (set auto_vlan=no) doesn't work with bnx driver.

Comment 3 Lior Vernia 2015-03-04 14:01:42 UTC
Hi Pavel,

First of all, this bug was reported on 3.4 (not sure which z-stream version), so changed the version accordingly. In fact, in 3.5 this might be fixed.

This definitely can't work on versions before 3.5 when the interface name isn't of format "<nic-name>.<vlan-id>" - it is indeed the "-fcoe" suffix that screws things up.

I'm not very knowledgeable about FCoE, is it possible to simply rename the ifcfg file (I assume they're originally created by a FCoE command/daemon), update the ifcfg file's contents and restart the network service? Would FCoE remain functional?

Again, in 3.5 I'd expect things to work (once the bond issue is resolved).

Comment 4 Pavel Zhukov 2015-03-04 14:25:34 UTC
(In reply to Lior Vernia from comment #3)
> Hi Pavel,
> 
> First of all, this bug was reported on 3.4 (not sure which z-stream
> version), so changed the version accordingly. In fact, in 3.5 this might be
> fixed.

vdsm-4.14.18-4.el6ev.x86_64  
rhevm-3.4.4-2.2.el6ev.noarch   

> 
> This definitely can't work on versions before 3.5 when the interface name
> isn't of format "<nic-name>.<vlan-id>" - it is indeed the "-fcoe" suffix
> that screws things up.
> 
> I'm not very knowledgeable about FCoE, is it possible to simply rename the
> ifcfg file (I assume they're originally created by a FCoE command/daemon),
> update the ifcfg file's contents and restart the network service? Would FCoE
> remain functional?
> 
No, it's not possible. ifcfg file itself doesn't contain the suffix. It's being added by fcoemon program which is binary (some kind of analog of ifup script).

Example:
# cat ifcfg-em1_1 
HWADDR="00:0A:F7:66:ED:B0"
ONBOT=yes
PEERDNS=no

result:
# ifconfig -a 
em1_1.1001-fcoe Link encap:Ethernet  HWaddr 00:0A:F7:66:ED:B0  
          inet6 addr: fe80::20a:f7ff:fe66:edb0/64 Scope:Link
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:0 errors:0 dropped:0 overruns:0 frame:0
          TX packets:0 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0 
          RX bytes:0 (0.0 b)  TX bytes:0 (0.0 b)



> Again, in 3.5 I'd expect things to work (once the bond issue is resolved).
I don't think so (I can be mistaken of course. can you please elaborate a little which commit is supposed to fix this?). However I don't have FCoE infrastructure to test it.

Comment 5 Lior Vernia 2015-03-04 15:11:24 UTC
Bug 999975 was fixed in 3.5, allowing for arbitrarily-named VLAN devices to be correctly displayed in the engine.

If this indeed turns out to be fine in 3.5, will the 3.4 issue still be relevant? Or should everything be upgraded to 3.5?

Comment 6 Pavel Zhukov 2015-03-05 08:26:55 UTC
Thank you for pointing this. Closing as duplicate

*** This bug has been marked as a duplicate of bug 999975 ***


Note You need to log in before you can comment on or make changes to this bug.